> but I'd easily make a bet that in 5 years that LLMs will be nowhere near human-level intelligence on some tasks

I wouldn't. There are some extraordinarily stupid humans out there. Worse, making humans dumber is a proven and well-known technology.